YOLOv7s Optimization for Road Defect Detection: Pruning, Pooling and Attention Mechanisms

Ediga Nidiganti Rishika Thanmai

2025

Abstract

Road defect detection is crucial for ensuring traffic safety and efficient infrastructure maintenance. This report presents an optimized YOLOv7-based road defect detection system, integrating pruning, pooling, and attention mechanisms to enhance accuracy, reduce computational complexity, and improve real-time performance. Pruning techniques eliminate redundant parameters, accelerating inference speed to maintain detection accuracy. Pooling strategies, including Spatial Pyramid Pooling (SPP) and Adaptive Pooling, enhance multi-scale feature extraction, enabling the model to detect defects of various shapes and textures. Additionally, attention mechanisms such as the Convolutional Block Attention Module refine feature selection, focusing on critical defect regions and reducing false positives. Experimental results demonstrate that the proposed optimizations significantly improve precision, recall, and mean average precision (mAP) on benchmark datasets while minimizing computational overhead. The enhanced YOLOv7 model is lightweight and efficient, making it ideal for real-time road monitoring applications, smart city infrastructure, and autonomous vehicle systems.

Paper Citation

in Harvard Style

Thanmai E. (2025). YOLOv7s Optimization for Road Defect Detection: Pruning, Pooling and Attention Mechanisms. In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25; ISBN 978-989-758-777-1, SciTePress, pages 707-712. DOI: 10.5220/0013904200004919

in Bibtex Style

@conference{icrdicct`2525,
author={Ediga Thanmai},
title={YOLOv7s Optimization for Road Defect Detection: Pruning, Pooling and Attention Mechanisms},
booktitle={Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25},
year={2025},
pages={707-712},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013904200004919},
isbn={978-989-758-777-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25
TI - YOLOv7s Optimization for Road Defect Detection: Pruning, Pooling and Attention Mechanisms
SN - 978-989-758-777-1
AU - Thanmai E.
PY - 2025
SP - 707
EP - 712
DO - 10.5220/0013904200004919
PB - SciTePress